this post was submitted on 12 Nov 2023
88 points (83.8% liked)
Asklemmy
43516 readers
2163 users here now
A loosely moderated place to ask open-ended questions
Search asklemmy ๐
If your post meets the following criteria, it's welcome here!
- Open-ended question
- Not offensive: at this point, we do not have the bandwidth to moderate overtly political discussions. Assume best intent and be excellent to each other.
- Not regarding using or support for Lemmy: context, see the list of support communities and tools for finding communities below
- Not ad nauseam inducing: please make sure it is a question that would be new to most members
- An actual topic of discussion
Looking for support?
Looking for a community?
- Lemmyverse: community search
- sub.rehab: maps old subreddits to fediverse options, marks official as such
- !lemmy411@lemmy.ca: a community for finding communities
~Icon~ ~by~ ~@Double_A@discuss.tchncs.de~
founded 5 years ago
MODERATORS
you are viewing a single comment's thread
view the rest of the comments
view the rest of the comments
I am not yet tired of pointing out that you cannot rely on a LLM for facts. LLMs like ChatGPT have no concept of what a fact is. Unless you already know something about a topic you won't even notice when they are making stuff up.
It's good to be cautious. I agree. Indeed, as I have deep expertise in programming, I recognize when it is over-complicating things or outright hallucinating. And I'll double check output when it's important.
But that doesn't discount the incredible usefulness of these tools. I've noticed a 20-30% productivity boost in my work, and googling for things now feels like a step back to the dark ages. Stack Overflow laid off 28 percent of its staff.
Even as just a sounding board, mentor, coach, and idea generator, the tools are so helpful. And that aspect doesn't require 100% accuracy. And if you think about it, it's not like pre-LLM documentation on the web or chatting with colleagues was ever flawless. We've all run into a post online that was confidently wrong, or a coworker that stubbornly insisted on something stupid.
Here's another thing to consider: while the tools have flaws and limits, this is the worst they'll ever be going forward. There's constant new improvements, like like tree of thought prompting and multi-modality. Just the other day, I took a photo of a wire mess near my home router with GPT Vision, and I had the LLM suggest cable-neatening products and methods (I've always struggled with cable management).
In fact, the biggest limit I've noticed is simply people's lack of creativity in using the tools (or willingness to use them), not the tools themselves.
They're here to stay. ChatGPT became the most-used / most-quickly adopted product of all time for a reason. Those who are willing to work with them and learn about them (both their strengths and weaknesses) will benefit, and those unwilling to do so, and who are too-dismissive, will fall behind.
Oh, yes, they absolutely have their uses. But at least at the moment they don't live up to the hype. And until they get their facts straight, they are mostly useless for my job. I do keep warning people because I do see a worrying trend of people assuming ChatGPT actually knows what it's talking about. And that is not the case.
Some do cite their sources
Yeah, and the sources are sometimes made up